metadata

language:
  - tr
task_categories:
  - visual-question-answering
size_categories:
  - 100K<n<1M

Dataset Card for ImageText-595K-Turkish

Language: Turkish

Dataset Description

This repository contains a dataset used for pretraining the Turkish version of the JanusPro model. The dataset is a Turkish translation of the English dataset previously used in similar studies. The translation was performed using the DeepL translation model to ensure high-quality, accurate translation.

Dataset Details

Size: ≈600K
Data source: The original dataset was based on [liuhaotian/LLaVA-CC3M-Pretrain-595K] and was adapted to Turkish.
Translation tool: DeepL
Data format: Images, Prompt